
Traditional
Traditional
Traditional
Traditional
Multi-System
Multi-System
Multi-System
Multi-System
High Availability
High Availability
High Availability
High Availability
49
To configure a cluster with no SPOF, the membership must extend beyond a single cabinet. The
cluster must be configured such that the failure of a single cabinet does not result in the failure
of a majority of the nodes in the cluster. The cluster lock device must be powered independently
of the cabinets containing the cluster nodes. Alternative cluster lock solution is the Quorum
Service, which resides outside the Serviceguard cluster providing arbitration services.
50
A cluster lock is required if the cluster is wholly contained within two single cabinets (i.e., two
Superdome/16-socket or 32-socket systems or two Superdome/PA-8800 32-socket or 64-
socket systems) or two dual cabinets (i.e. two Superdome/64-socket systems or two
Superdome/PA-8800 128-socket systems). This requirement is due to a possible 50% cluster
failure.
51
Serviceguard only supports cluster lock up to four nodes. Thus a two cabinet configuration is
limited to four nodes (i.e., two nodes in one dual cabinet Superdome/64-socket system or
Superdome/PA-8800 128-socket system and two nodes in another dual cabinet
Superdome/64-socket system or Superdome/PA-8800 128-socket system). The Quorum
Service can support up to 50 clusters or 100 nodes (can be arbitrator to both HP-UX and Linux
clusters).
52
Two-cabinet configurations must evenly divide nodes between the cabinets (i.e. 3 and 1 is not a
legal 4-node configuration).
53
Cluster lock must be powered independently of either cabinet.
54
Root volume mirrors must be on separate power circuits.
55
Redundant heartbeat paths are required and can be accomplished by using either multiple
heartbeat subnets or via standby interface cards.
56
Redundant heartbeat paths should be configured in separate I/O chassis when possible.
57
Redundant paths to storage devices used by the cluster are required and can be accomplished
using either disk mirroring or via LVM's pvlinks.
58
Redundant storage device paths should be configured in separate I/O chassis when possible.
59
Dual power connected to independent power circuits is recommended.
Heterogeneous
Heterogeneous
Heterogeneous
Heterogeneous
Multi System
Multi System
Multi System
Multi System
High Availability
High Availability
High Availability
High Availability
60
Cluster configurations can contain a mixture of Superdome and non Superdome nodes.
61
Care must be taken to configure an even or greater number of nodes outside of the Superdome
cabinet
62
If half the nodes of the cluster are within a Superdome cabinet, a cluster lock is required
(4-node maximum cluster size)
63
If more than half the nodes of a cluster are outside the Superdome cabinet, no cluster lock is
required (16-node maximum Serviceguard cluster size).
64
Up to a 4-node cluster is supported within a single cabinet system (Superdome/16-socket or
Superdome/PA-8800 32-socket)
65
Up to an 8-node cluster is supported within a single cabinet system* (Superdome/32-socket or
Superdome/PA-8800 64-socket)
66
Up to a 16-node cluster is supported within a dual cabinet system* (Superdome/64-socket or
Superdome/PA-8800 128-socket)
67
Cluster lock is required for 2-node configurations
68
Cluster lock must be powered independently of the cabinet.
69
Root volume mirrors must be on separate power circuits.
70
Dual power connected to independent power circuits is highly recommended.
QuickSpecs
HP Integrity Superdome Servers 16-socket, 32-socket,
HP Integrity Superdome Servers 16-socket, 32-socket,
HP Integrity Superdome Servers 16-socket, 32-socket,
HP Integrity Superdome Servers 16-socket, 32-socket,
and 64-socket
and 64-socket
and 64-socket
and 64-socket
Configuration
DA - 11717 Worldwide — Version 20 — May 3, 2005
Page 37